Paul Bush trills:
> In article <4p22p2$ihf at nntp1.best.com>, Sky Coyote <sky at intergalact.com> writes:
> |> >These are no problem (what do you mean by hyperaccurate? Very detailed?)
> |>
> |> I mean producing exactly the same output as the real thing, rather than
> |> just qualitatively similar output. Ideally, the time-series difference
> |> between the neuron (or neurons) and the simulation should be less than
> |> a given epsilon for all t > 0, for all relevant metrics (e.g. electro-
> |> chemical potential, etc...).
>> I think modelers aim for these ideals already. The problem comes in defining the
> 'relevant metrics' and in accumulating enough high precision data that you can
> trust to constrain your simulation. Biology is very variable, biological data as
> reported in scientific journals even more so.
I agree. We are more interested here in trying to model slightly larger-scale nets so
Genesis is a suitable low-level tool for the single-neuron approach. If there are any
commercially available modelling tools that go into more detail than the Hodgkin-Huxley
equation I haven't heard of them, but that's not to say you can't develop your own
(you'll probably have to end up doing this in any case since you'll always find that
any standardised modelling tool leaves out the interesting parts you feel are important
in the model).
dog
--
mailto:dog at cogsci.ed.ac.ukhttp://www.cogsci.ed.ac.uk/~dog
"Time for a less subtle approach." - Kira Nerys
